10 ms·
Neural networks do not think
by silent_cal 3y ago
Neural networks do not think
- adlpz 3y agoDo neurons think? Do a bunch of neurons? Is this semantics?
- empath-nirvana 3y agobasically this: https://en.wikipedia.org/wiki/Sorites_paradox https://en.wikipedia.org/wiki/Sorites_paradox One neuron doesn't think. Three neurons don't think. Billions of neurons think. Somewhere between one neuron and billions of neurons, thinking starts happening. Probably also true for neural networks. The main problem is that people throw around terms like: "Thought", "Intelligence", "Will", "Reasoning", "Knowledge", "Consciousness", etc like they are very well defined and well understood terms and they very much are not.
- silent_cal 3y agoBillions of neurons don't think, people do.
- empath-nirvana 3y ago...with what?
- silent_cal 3y agoWith their minds
- adlpz 3y agoMy point precisely. Those are all vague terms. Saying that "neural nerworks do not think" is as meaningless as any equivalent (or opposite) statement on any other system including any number of neurons, a whole brain or a person. It's all semantics.
- silent_cal 3y agoYour claim is that saying "people think" is meaningless? Maybe you don't think (seems to be the case), but I certainly do.
- adlpz 3y agoIt is meaningless to me because the term is imprecise. Is it precise when I do something bone headed and say "sorry I wasn't thinking"? Do cats think? Do worms? Depends on what you mean by thinking in that particular case. In my opinion what LLMs do is close enough to thinking for some definition of the word. And, by the way, I think there's no need to be unpleasant to state your point.
- silent_cal 3y agoThere's a difference between a term being imprecise and being equivocal. Yeah, we use the term "think" in different ways. But it primarily refers to the operations unique to the human intellect with which we're all familiar.. deduction, analysis, synthesis, judgment, deliberation, abstraction, etc. The thing that's special about these operations is that they deal with immaterial concepts. What's right? What's good? What's true? And so on. Computers have never achieved this and they never will, because they are only material arrangements, and so can only handle things on the material level. The human mind is clearly not reducible to that, because it can handle purely immaterial concepts.
- adlpz 3y agoI see where we differ in opinion. I do believe that the human mind is just a very particular material arrangement. Therefore while different in scale to an LLM and likely also differing in structure, I do think both systems are fundamentally in the same category. Given that, we could never agree. But that's fine. And by the way, I actually hope you're right.
- silent_cal 3y ago
- silent_cal 3y agoThere are no real neurons in a neurons in a neural network.
- PoignardAzur 3y agoYou're not giving information anybody on this forum doesn't already know. Obviously they don't "speak" either. Both "think" and "speak" are used as shorthands here for what the language models actually do.
- silent_cal 3y agoWhat are you upset with me for? The authors are using the misleading language, not me. Take it up with them.
- optimalsolver 3y agoCould you give a definition of "think" that NNs fail to live up to?
- silent_cal 3y agoAbstracting immaterial concepts from physical reality and deliberately using them in analytical or deductive processes to discover truths.
- optimalsolver 3y agoSo basically finding ways to compress your observational history?
- silent_cal 3y agoNo, it's not "basically" that at all.
- optimalsolver 3y agoThat's pretty much what it is, as you stated it. Finding abstractions that let you encode your observational history more efficiently than you previously could, or "discovering truths", if you want to be all mystical about it.
- silent_cal 3y agoThat's not what it is, and it's not what I stated.
- optimalsolver 3y agoThen could you go into more detail? Because what I just described was the progress of scientific theory. If an AI can do that, it's not going to matter whether or not it meets your arcane definition of "thinking".
- 4RealFreedom 3y agoI don't understand the downvotes - you are correct.
- silent_cal 3y agoI think people just get mad when they're reminded of this obvious fact. They want computers to prove that our minds are an illusion, the product of a "meat computer".
- stoniejohnson 3y agoRead some Daniel Dennett!
- silent_cal 3y agoAre you serious?
- stoniejohnson 3y agoYou're very grumpy I think you need some food and a nap :-)
- silent_cal 3y agoI think you need religion
- viraptor 3y agoIt's fair to talk about thinking in a handwavey "you know what I mean" way. This is not a philosophy paper. It's a fine point if that's what you want to discuss, but doesn't change anything about the issue at hand and is needlessly pedantic. It's the "what you're referring to is actually GNU/Linux" of discussions about the tech side of AI.
- silent_cal 3y ago